Newer milestone thread - Page 88

bloodIce · 2013-12-04, 09:29

Congratulations for the factor. It seems that there are some of those left

.
As for Factor5, in my hands if I run it on 4cores it is about 2,5x slower than Primer95 run on one core (so in total about 10x slower). If I compare it to mfaktc it will be even worse. It is an nice idea to utilize all cores for an expo, but Primer95 is still better. I might missed some setting though.

ET_ · 2013-12-04, 10:52

Quote:

Originally Posted by bloodIce

Congratulations for the factor. It seems that there are some of those left

.
As for Factor5, in my hands if I run it on 4cores it is about 2,5x slower than Primer95 run on one core (so in total about 10x slower). If I compare it to mfaktc it will be even worse. It is an nice idea to utilize all cores for an expo, but Primer95 is still better. I might missed some setting though.

You are right, I forgot that my timings referrerd to Prime95 not multi-threaded, so that 16 threads of Factor5 were a bit faster than Prime95 single thread. I still may have some ideas for optimization, but they won't compare to the new Prime95.

We defiitely need a customized CUDA program for small expos...

Luigi

lycorn · 2013-12-04, 12:28

Just reserved a batch of 20 (2.51M,62,63).
I´m using an old Intel dual Core @1.8GHz. Not working 24/7, so it will take a couple of days.

diamonddave · 2013-12-04, 14:13

Quote:

Originally Posted by lycorn

Me, myself and I: 2.52M,62,63.
I noticed David Campeau has been actively working on 7M,64,65.
If everybody cares to reserve the expos and to only take unreserved, no stepping will occur.

Yeah I use Prime95 to reserve unassigned expo in the 1M - 10M range that are at 64 bit. Then I switch my worktotdo file to mfaktc. If everyone reserve with Primenet, no one should step on any toes.

I'm also work at doing P-1 in that range (unfortunatly Primenet doesnt track those reservation). If anyone want to jump in to help there, we will have to coordinate the effort. To date I have found 312 factors in the last 3 months. I'm adding 5% more factor probability to the existing work, but I only get 2.3% success rate thus far. Probably because there was a lot of ECM work done in the range.

LaurV · 2013-12-04, 14:25

Quote:

Originally Posted by lycorn

Hmmm... not quite. IIRC, the "Less Classes" variant was recommended for assignments of the type 300/400M+ from 65 to 66, or something of that sort, because those ones yes, they are really short.
Taking, say, 1.1M from to 62 to 63 bits is not a short assignment, so I don´t think it qualifies for the use of that particular version.

Correct. I tried it and you are right. Sorry

BTW, finished my range, 8 7 (sorry, wrong counting! one was from a different range) factors found.

Code:

M2186747 has a factor: 6816375925079296039 [TF:62:63*:mfaktc 0.20 75bit_mul32]
M2192387 has a factor: 4772503165120678223 [TF:62:63*:mfaktc 0.20 75bit_mul32]
M2232889 has a factor: 7490062285023913423 [TF:62:63*:mfaktc 0.20 75bit_mul32]
M2202533 has a factor: 4822443047224966873 [TF:62:63*:mfaktc 0.20 75bit_mul32]
M2246551 has a factor: 8939019228125586599 [TF:62:63*:mfaktc 0.20 75bit_mul32]
M2206759 has a factor: 8117337983834625703 [TF:62:63*:mfaktc 0.20 75bit_mul32]
M2212433 has a factor: 9227837650261017119 [TF:62:63*:mfaktc 0.20 75bit_mul32]

I will not work in the 2M-3M range for the time being.

blahpy · 2013-12-04, 19:55

Quote:

Originally Posted by blahpy

In order to clear up another range I've quickly written up a small Java program to convert factoring effort reports into mfaktx lines, so I'll get a set of lines to clean up 800M - 900M to 66 (even though it's pretty pointless work, that range is really close to being done but the leftovers that are only TFed to 65 are sparse and OCD says it should be fixed).

If someone wants the lines when I have them (Mawn perhaps?), I could quite easily send them over, there are only 6,000 or so and they will take probably 5 or so seconds per exponent on a GTX 670 or whatever you guys use.

I have finished tidying it up.

ET_ · 2013-12-04, 20:19

Quote:

Originally Posted by blahpy

I have finished tidying it up.

Are you using GPU sieving? If so, I can take 1,000 exponents while a new list for low-factored 2M exponents comes out.

Luigi

blahpy · 2013-12-04, 20:27

Quote:

Originally Posted by ET_

Are you using GPU sieving? If so, I can take 1,000 exponents while a new list for low-factored 2M exponents comes out.

Luigi

I meant, I have taken those last 6,000 exponents up to 2^66 myself already :)

ET_ · 2013-12-04, 20:33

Quote:

Originally Posted by blahpy

I meant, I have taken those last 6,000 exponents up to 2^66 myself already :)

Luigi

flashjh · 2013-12-04, 21:16

Quote:

Originally Posted by bloodIce

Any hope for those ~3,2K low-hanging fruits from 2M to 2.5M, which need a push from 62 to 63 bits http://www.mersenne.info/trial_factored_tabular_data/3/2000000/? I am sure some factors will pop up. With your resources this will be a half day work. I occasionally do several expos in this range, but it will be nice if the heavy guns shoot once.

Quote:

Originally Posted by ET_

Correct! How may exponents (so that I will update the remainders)?

Luigi

All done. Found 12 factors

Code:

M2353657 has a factor: 9158731620188823601 [TF:62:63*:mfaktc 0.20-LessClass(flashjh) 75bit_mul32]
M2357167 has a factor: 6862040924454446641 [TF:62:63*:mfaktc 0.20-LessClass(flashjh) 75bit_mul32]
M2357483 has a factor: 7205650442358613943 [TF:62:63*:mfaktc 0.20-LessClass(flashjh) 75bit_mul32]
M2380241 has a factor: 8948578016786415841 [TF:62:63*:mfaktc 0.20-LessClass(flashjh) 75bit_mul32]
M2386859 has a factor: 7761331228334735449 [TF:62:63*:mfaktc 0.20-LessClass(flashjh) 75bit_mul32]
M2403913 has a factor: 5252674518907474873 [TF:62:63*:mfaktc 0.20-LessClass(flashjh) 75bit_mul32]
M2410351 has a factor: 6458874527758286167 [TF:62:63*:mfaktc 0.20-LessClass(flashjh) 75bit_mul32]
M2420417 has a factor: 6353821671263040703 [TF:62:63*:mfaktc 0.20-LessClass(flashjh) 75bit_mul32]
M2422781 has a factor: 5029481903039574241 [TF:62:63*:mfaktc 0.20-LessClass(flashjh) 75bit_mul32]
M2423581 has a factor: 4801187570656498871 [TF:62:63*:mfaktc 0.20-LessClass(flashjh) 75bit_mul32]
M2451433 has a factor: 5573887011291003313 [TF:62:63*:mfaktc 0.20-LessClass(flashjh) 75bit_mul32]
M2484739 has a factor: 6449682757245875783 [TF:62:63*:mfaktc 0.20-LessClass(flashjh) 75bit_mul32]

bloodIce · 2013-12-04, 22:28

Amazing work guys

(I intentionally do not mention names because even those TFing one or two expos contribute, but special thanks to the big guns)! Together for less then 48 hours you did around 2,500 of those expos. I cannot believe it. There are some 625 left (http://www.mersenne.org/report_factoring_effort/?exp_lo=2000000&exp_hi=2600000&bits_lo=0&bits_hi=62&exassigned=1&B1=Get+Data). And some 74 reserved, but the range is almost done.You found at least 12+7+1 = 20 factors. Yeah, not what you dreamt for, but around 0.8% is not that bad for a Pm1ed and slightly ECMed range. Most importantly, soon there will be no 62bit expos left (above the million). So at the and of the range we expect around 10 new factors.
I slowly progress with my >M6,666,000 to 64bits as I promised. I realised that those below 1M are not completely lost cause, but very hard (and this will be primary Prime95 job).

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Newer X64 build needed	Googulator	Msieve	73	2020-08-30 07:47
Performance of cuda-ecm on newer hardware?	fivemack	GMP-ECM	14	2015-02-12 20:10
Cause this don't belong in the milestone thread	bcp19	Data	30	2012-09-08 15:09
Newer msieves are slow on Core i7	mklasson	Msieve	9	2009-02-18 12:58
Use of large memory pages possible with newer linux kernels	Dresdenboy	Software	3	2003-12-08 14:47

2013-12-04, 09:29	#958
bloodIce Feb 2010 Sweden 10101101₂ Posts	Congratulations for the factor. It seems that there are some of those left . As for Factor5, in my hands if I run it on 4cores it is about 2,5x slower than Primer95 run on one core (so in total about 10x slower). If I compare it to mfaktc it will be even worse. It is an nice idea to utilize all cores for an expo, but Primer95 is still better. I might missed some setting though.

2013-12-04, 12:28	#960
lycorn "GIMFS" Sep 2002 Oeiras, Portugal 2·11·67 Posts	Just reserved a batch of 20 (2.51M,62,63). I´m using an old Intel dual Core @1.8GHz. Not working 24/7, so it will take a couple of days.

2013-12-04, 22:28	#968
bloodIce Feb 2010 Sweden 10101101₂ Posts	Amazing work guys (I intentionally do not mention names because even those TFing one or two expos contribute, but special thanks to the big guns)! Together for less then 48 hours you did around 2,500 of those expos. I cannot believe it. There are some 625 left (http://www.mersenne.org/report_factoring_effort/?exp_lo=2000000&exp_hi=2600000&bits_lo=0&bits_hi=62&exassigned=1&B1=Get+Data). And some 74 reserved, but the range is almost done.You found at least 12+7+1 = 20 factors. Yeah, not what you dreamt for, but around 0.8% is not that bad for a Pm1ed and slightly ECMed range. Most importantly, soon there will be no 62bit expos left (above the million). So at the and of the range we expect around 10 new factors. I slowly progress with my >M6,666,000 to 64bits as I promised. I realised that those below 1M are not completely lost cause, but very hard (and this will be primary Prime95 job).